Link prediction is a very well studied problem as it has applications in different areas. Many algorithms have been presented in the literature for link prediction problem. The link prediction problem can be explained as follows: given the topology of graph G at a certain time t, we need to predict the topology of the graph at time t′ where t′ > t. The techniques used for link prediction are categorized as follows: nodes based techniques, Link based techniques and path based techniques. There are some other techniques that use meta-approaches. In this paper we present a new algorithm for link prediction on social networks. The algorithm tries to identify nodes that may get deleted by time t′ and use this information to predict new links that might appear in the future. Our tests show that our algorithm shows better result for link prediction on social networks compared to the previous algorithms. We had implemented our algorithm in C# on Pentium Core 2 Duo processor. The data used for testing is Gnutella peer to peer network. © 2014, The Society for Reliability Engineering, Quality and Operations Management (SREQOM), India and The Division of Operation and Maintenance, Lulea University of Technology, Sweden.